VOLUME

JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR

7,

NUMBER 4

JULY, 196;4

FIXED-RATIO PUNISHMENT
WITH CONTINUOUS REINFORCEMENT'
DEREK P. HENDRY AND C. VAN-TOLLER
DURHAM UNIVERSITY
Rats were reinforced with water for every bar-press and concurrently punished for every 10th
or 20th bar-press. Punishment produced an initial suppression of responding followed by
recovery. A slight change in the method of delivering punishment eventually led to a high
response rate just after punishment, a low response rate just before punishment, and frequent
intermediate pauses. The results are interpreted as showing that punishment became a safe
signal and that the high rate of responding it released came to act as a conditioned aversive
stimulus. The effects of amphetamine were consistent with this interpretation. Alcohol had
the paradoxical effect of increasing pauses and depressing the low rate before punishment.

Azrin (1956) first suggested that the delivery

of aversive stimuli according to different schedules might produce typical patterns of responding related to those produced by the
corresponding schedules of reinforcement. He
maintained responding on a variable-interval
schedule of reinforcement (VI reinf) and introduced signaled periods in which punishment was also scheduled. The complete schedule, in the case of fixed-interval punishment
(Fl pun), could be represented as Mult [VI
reinf; Con (VI reinf, Fl pun)]. Azrin interpreted the relative changes in response rate
during "dangerous" periods as supporting the
general view that punishment has the opposite
effect from reinforcement. With Fl pun, for
example, rate was negatively accelerated.
With fixed-ratio punishment, response rate
might also be expected to decline as the number of unpunished responses increased. This
result was not obtained by Azrin, Holz, and
Hake (1963), using concurrent VI reinf and
FR pun. They found a uniform depressant
effect of punishment on response rate, and
only rare instances of a differentiated pattern
of responding between punishments. However,
VI reinf differentially reinforces low rates.
Consequently, the reinforcement schedule
would counteract any tendency of the punishment schedule to produce high rates. This
consideration is less relevant to Azrin's original study of response-correlated shock, in

which periods of punishment were explicitly

signaled and punishments could be avoided
at very small cost in reinforcements. It was
because of such possible interaction that we
chose continuous reinforcement (CRF), which
does not differentially reinforce rate, as the
concurrent reinforcement schedule with FR
pun.
METHOD

Subjects
Eight male albino rats, about 3 months old
at the beginning of the experiment, served.

Apparatus
Three identical operant conditioning boxes,
with control and recording equipment located
in an adjacent room were used. The boxes
measured 8 in. by 9 in. by 10 in. high, with
aluminum walls and a clear plastic top. The
floor was a grid consisting of seven alloy rods,
5/8 in. in diameter, mounted on 11/8 in. centers,
running the length of the box. A bar, operated
by a force of 30 g, projected from one (narrow)
wall, 3 in. above the floor. A small drinking
well was located just below and to one side of
the bar. Electric shock from the secondary of
a power transformer could be applied either to
alternate rods of the grid ("grid-shock") or
between the entire grid and the bar ("barshock"). Thus, grid shock was to the hind
paws, and bar shock was between fore paws
'Reprints may be obtained from Derek P. Hendry,
Space Research Laboratory, Dept. of Psychology, Uni- and hind paws. Shock intensity was controlled
by a potentiometer and limited by a 220 K
versity of Maryland, College Park, Maryland.

293

294

DEREK P. HENDRY and C. VAN-TOLLER

Fig. 1. Typical initial effect of fixed-ratio punishment (300 v), and later recovery; S5 on the left, S4 on the right.
Every response was reinforced and pips indicate punished responses. Session numbers are circled;-I indicates the
pre-punishment session. Punishment was omitted on session 42.

fixed-ratio punishment. They were trained to

bar-press for water and given 200 reinforcements in one session, before the administration
of punishment. The amount of water per
reinforcement was 0.03 ml. Reinforcements, up
to the capacity of the drinking well (about 0.1
Procedure
ml.), remained available until consumed. SubTwo of the subjects (S7 and S8) were ex- jects were given water ad lib for 15 min shortly
posed only to continuous reinforcement (CRF), after each daily session, which generally lasted
limited to 200 reinforcements per session. Most 1 hr.
of the results were obtained from the other six
Punishment (300 v grid-shock-i.e. no shock
Ss (SI to S6), which received both CRF and on the bar) was given on every 20th response
ohm series resistor. Shock duration was about
50 msec. Each box had a water pump, which
operated with slight vibration and noise. Consequently, distinctive exteroceptive feedback
was normally provided for every response.

FIXED-RATIO PUNISHMENT

295

punishment was followed by long pauses and

positively accelerated responding. At a later
stage, when pauses occurred, they tended not
to appear immediately after shock. Eventually
the rate of responding seemed to be little
affected by punishment. The fully recovered
response rate was about the same as it had
been during the session of CRF and the only
effect of omission of punishment was a greater
persistence of responding towards the end of
the session. All these effects are illustrated in
Fig. 1. No effect could be detected of the
change to FR 10, which occurred after complete recovery of bar-pressing.

Fig. 3. Short-term effect of increasing punishment

intensity. The upper trace shows performance with a
punishment intensity of 300 v. The lower trace shows
performance on the next session with a punishment
intensity of 500 v. Pips indicate punished responses.

(FR 20). The shock went on the instant the

microswitch of the bar operated. After 18 sessions, the schedule was changed to FR 10 for
three Ss. After performance had stabilized,
punishment was omitted for one session. BarFig. 4. S6. Short-term effect of increasing punishment
shock was then substituted for grid-shock for
intensity in successive sessions.
the remainder of the experiment-some 100 A:
Punishment intensity 300 v.
sessions. During these sessions the intensity of B: Punishment
intensity 500 v.
punishment was changed from time to time. Pips indicate punished responses.
The effects of dl-amphetamine sulfate and
ethyl alcohol on the performance of all eight
Ss were studied. The experimental animals
were given i/p injections immediately before
a session. Test sessions with drugs were separated by a number of sessions with injections
of the vehicle (physiological saline or distilled
water). Drugs were not administered unless
performance had returned to the pre-drug
baseline on the preceding (control) session.
The amphetamine doses were 1.0, 2.0 and 3.0
mg/kg. The alcohol doses were 0.5, 1.0, 1.5
and 2.0 mg/g. The weight of all injections was
1% of the animal's body weight. The effect of
drinking 8 ml water immediately before a sesFig. 5. Short-term vs long-term effects of increasing
sion was also determined.
punishment intensity; SI on the left, S4 on the right.

RESULTS
Grid-Shock
The first few shocks suppressed bar-pressing,
followed by a gradual recovery which was complete after about 15 sessions. During recovery,

Pips indicate punished responses. Session numbers are

circled. They represent the number of bar-shock sessions, not the number of sessions since the beginning of
the experiment. The punishment intensity was 300 v
on session 13 and 400 v on sessions 14 through 16. SI
shows a partial recovery of original performance at the
lower intensity; S4 shows a progressive decline in performance.

DEREK P. HENDRY and C. VAN-TOLLER

A

A
100

AA

40

70

. -

It

WI0/

Cy

/ I

10m

CI

75f

A
110

10min
10min

A+9Omln

Fig. 6. The effect of amphetamine on fixed-ratio

punishment (FR 10). Performance after: A, saline i/p;
B, 1.0 mg/kg amphetamine i/p; C, 3.0 mg/kg amphetamine i/p. Pips indicate punished responses. S3 was
replaced in his home cage for 90 min between the two
records shown as "A" and "A and 90 min." Numbers
indicate the length in minutes of a portion of the
record extracted at the break. Punishment intensity
was 150 v for SI (top), 300 v for S2 (middle) and 400 v
for S3 (bottom).

Bar-Shock
The change to bar-shock produced a gradual
and irregular decline and partial recovery in
rate of bar-pressing, requiring 6 to 30 sessions
depending on the S. The rate of bar-pressing
even' after recovery was usually lower than
with grid-shock. In addition, performance
showed slow drifts throughout the rest of the
experiment.
Differentiated response rates between punishments emerged with bar-shock. The basic
pattern was a high rate immediately after
punishment, followed by a low rate, sometimes
accelerating, up to the next punishment. The
two rates were often separated by relatively
long pauses or very low rates, as shown in
Fig. 2. The records shown in Fig. 2 were selected to show differentiation, since it was not

Fig. 7. The effect of amphetamine on fixed-ratio

punishment (FR 20). Punishment intensity was 200 v
for S4 (top), 350 v for S5 (middle), and 400 v for S6
(bottom). See Fig. 6.

always present. An upward drift in response

rates was usually accompanied by disappearance of the lower rate just before punishment.
A downward drift in response rates was accompanied by more frequent pausing and attenuation of the high rate just after punishment.
To maintain differentiation, the intensity of
punishment was raised or lowered whenever
the response rate was seen to have drifted upwards or downwards in the previous few sessions. Because of this procedure, few systematic data were obtained on the effects of punishment intensity. In view of the relative instability of performance, any attempt to relate a
fixed level of performance to a given punishment intensity would be misleading. However,
changes in punishment intensity did produce
changes in performance. Figure 3 shows that,
except for the doubtful case of S4, the overall
response rate is lowered by an increase in the
punishment intensity from 300 v to 500 v.
If the response rate was high, so that differentiation was weak, an increase in punishment
intensity usually increased differentiation.
This is illustrated in Fig. 4.

FIXED-RA TIO PUNISHMENT

A'

297

B
B

Fig. 8. The effects of alcohol and satiation on fixed-ratio punishment (FR 10). Performance after: A, distilled
water i/p; B, 0.5 mg/g alcohol i/p; C, 2.0 mg/g alcohol i/p; D, drinking 8 ml of water immediately before the
session. Pips indicate punished responses. Numbers indicate the same as in Fig. 6. Records are from SI, S2, and S3,
reading from left to right. Punishment intensities are the same as in Fig. 6.

Figure 12 shows the effects of the drugs and

The initial effect of a change in intensity
might not be maintained in subsequent ses- pre-watering in detail in the case of SI and S6.
sions. Figure 5 shows examples of opposite It can be seen in Fig. 12 that amphetamine
effects after the initial lowering of response reduces the number of rapid responses emitted
rate by an increase in punishment intensity. just after punishment. In the case of S1, the
In one case the rate continues to decline; in rapid responding is eliminated. Alcohol and
the other it partially recovers.
pre-watering, on the other hand, have little
effect on the high rate after punishment. They
Effects of alcohol and amphetamine
appear mainly to increase pausing and reduce
Figures 6 to 9 show the drug effects for indi- the low rate just before punishment. These
vidual experimental Ss. Mean and median effects may also be seen in Fig. 6 to 9.
IRTs of all Ss are compared in Fig. 10 and
Fig. 11. It can be seen that the drug effects on
DISCUSSION
the performance of S7 and S8 are small in
The introduction of punishment had the
comparison with their effects on S1-S6. Figure 10 shows that pre-watering, alcohol and same effect as that found by Azrin et al. (1963).
amphetamine increase the mean IRT of Sl- Initial suppression was followed by the temS6. Figure 11 shows that amphetamine is the porary appearance of accelerated responding
only agent which increases median IRT. between Runishments. We obtained complete
Therefore, alcohol and pre-watering reduced recovery of initial response rate with the modthe overall rate of responding by increasing erate intensity of punishment used, whereas
the length, not the proportion, of long IRTs. Azrin et al., obtained complete recovery only
Amphetamine, on the other hand, increased at the lowest intensity of punishment.
The possibility was considered that the comthe proportion of long IRTs, as well, perhaps,
plete recovery of bar-pressing with grid-shock
as their length.

DEREK P. HENDRY and C. VAN-TOLLER

298

was due to the Ss learning to avoid most of

the shocks. This was extremely unlikely in
view of the design of the grid. Shock-avoidance
by standing on one rod or two rods of the
same polarity is very rare when the rods run
at right angles to the wall containing the bar.
The shock produced a distinct twitch in all
subjects. Frequent observation showed that
this sign of punishment was still present even
when no effect on response rate could be detected.
It is certain that systematic avoidance was
not learned by any S, but it is not certain that
every scheduled grid shock was received.
This was the reason for changing to bar-shock,
which would require very unusual circumstances to prevent the delivery of shock. Avoidance behavior such as operating the bar with
a furry part of the body was looked for but
never observed.
It was observed that the general bodily
twitch produced by shock was accompanied
by a recoil from the bar in the case of barshock, but not in the case of grid-shock. The
recoil followed every actual and abortive barpress for many sessions in the case of some Ss,

A
A.

being simply more vigorous when shock was

actually delivered. These observations are relevant to an "interference" theory of the effect
of punishment. According to this view, responses elicited by punishment are learned
and compete with the punished response.
Strong support for this view was obtained by
Fowler and Miller (1963). They showed that
a weak shock at the goal increased or decreased
runway speed according to whether the shock
was applied to the hind paws (eliciting forward motion) or the fore paws (eliciting backward motion). This result is similar to that
obtained in the present experiment. The effect
of the bar-shock (recoil from the bar) was
clearly incompatible with bar-pressing; but the
effect of grid-shock was not. Consequently,
bar-shock was a more effective deterrent than
the same intensity grid-shock.
The main result of this experiment is the
considerable amount of differentiated responding between punishments. Differences in this
respect between the present results and those
of Azrin et al. (1963) might be due to several
factors. We have already suggested that the
effects of differential reinforcement of rates

40

J
B

vFi

1Xe

ct

f
D

If

Fig. 9. The effects of alcohol and satiation on fixed-ratio punishment (FR 20). Records are from S4, S5, and S6,
reading from left to right. Punishment intensities are the same as in Fig. 7. See Fig. 8.

299

FIXED-RA TIO PUNISHMENT

(by a VI schedule) might conceal the effects

of differential punishment of rates. Another
possibility is that the crucial difference is in
the absolute rates of reinforcement or punishment. In both respects the two studies differ
by an order of magnitude.
The present results show that FR pun does
produce a recognizable pattern of behavior,
with perhaps more individual variation than
in the case of FR reinf. However, the pattern
of behavior produced by FR pun is not related
in a simple way to that produced by FR reinf.
The assumption that FR pun would have the
"opposite" effect from FR reinf was not confirmed.
Punishment came to act as a cue controlling
a relatively high probability of response. Since
all responses were reinforced, the punishment
did not act as a discriminative stimulus in its
usual sense of a signal for "not-extinction".
However, every punishment initiated a period
of unpunished responding. A punishment was
therefore a discriminative stimulus in the
novel sense of a signal for "no punishment".
Such a stimulus may be called a "safe signal".
200
MEAN & RANGE OF
MEANS S 1-S6

100-

in

TA

C7 IL CA

4 I.

i MEAN & RANGE OF

MEDIANS S1-S6.

*S7&S8.

6%7

z
uJ

I4
0

0 0

i0

1 0 2*0 3:0
1P0 15 2*0
AIc.(mglgm)
Amph.(mg/kg)
Fig. 11. Median IRTs calculated from the same sessions as those used in Fig. 10. The ordinate is linear.
The mean median IRT of SI-S6 was at least 9 sec
with the higher doses of amphetamine, but the recording apparatus was unable to resolve durations

PW

greater than 8 sec.

Whether a safe signal has all the properties

of an SD is an open question.
The low rate just before punishment may
50be regarded as a case of depression of posireinforced responses by cues which sig|Ttively
}
I.
nify that further responses may be punished
*
.. .
/
b
|(Hunt and Brady, 1955). Thus, in general,
1'
rapid responding gave way to slower responding. When the two response rates could be
*
< lo
X |
l
ljlclearly distinguished, the pattern of responding was obviously similar whether the punishment schedule was FR 10 or FR 20. Since no
*
punishment-correlated cues independent of
*
.0J
the Ss' behavior were provided, this result indi*
*
0 @
*
2
cates that response-produced cues mediated
@**
0
the control attributable to the punishment
schedule.
We interpret the results, therefore, as show,
,
,,W , ,~
20 3'-O
ing that FR punishments came to act as safe
s 1e 4 ;
ctw
410Ampb-(mgkg)
signals, releasing relatively rapid trains of
Ac.(mg/*
in
a
as
a
session
responses mainly under the control of the reinFig. 10. Means of the first 100 IRTs

l0
|.

JL

function of pre-watering 8 ml (PW), alcohol, and amphetamine. Control measures (C) were obtained from

Control sessions interspersed with the drug sessions.

The ordinate is logarithmic.

forcement schedule. These responses produced

conditioned aversive stimuli which in turn
depressed the initial high rate.

DEREK P. HENDRY and C. VAN-TOLLER

300

A~~~~~

B
A~~~~

D~~~~
D

10 min
Fig. 12. Detailed comparison of the effects of moderate doses of amphetamine and alcohol, and pre-watering. Traces on the left were produced by SI. Traces on
the right were produced by S6.
A, Typical control performance.
B, Performance after 2 mg/kg amphetamine.
C, Performance after 15 mg/g alcohol.
D, Performance after drinking 8 ml water before the
session. Pips indicate punished responses. All records
are of perfornance 30 min after the beginning of the
session.

The effects of amphetamine are consistent

with this analysis. Amphetamine is known to
depress responding during pre-aversive stimuli
(Brady, 1956). In addition, amphetamine reduces the rate of responding when responses
are both punished and reinforced (Geller and
Seifter, 1960). In short, amphetamine enhances
the effectiveness of aversive influences on reinforced behavior. In the present experiment
the high doses of amphetamine produced prolonged suppression of responding on the concurrent schedule, but not on CRF. Under
amphetamine, the train of rapid responses
after punishment was shorter; i.e., the transition from a high to a low rate, being due to
a competing aversive influence, was facilitated
by amphetamine.
Behavior suppressed by conditioned aversive
stimuli often recovers when the animal is
given alcohol (Conger, 1951; Masserman and
Yum, 1946; Miller and Barry, 1960). Alcohol
was therefore expected to alleviate the depres-

sion of responding just before punishment.

The effect was just the opposite-an enhancement of the depression of responding before
punishment. The effects of alcohol were similar to those of pre-watering. The possibility
that alcohol reduces the strength of all responses motivated by water deprivation cannot
be dismissed. Such an interpretation is implausible, however, since alcohol is a dehydrating agent. In addition, the results of the control Ss show that such an effect, if it exists, is
not strong enough to affect performance on
CRF.
The effects of alcohol do not fit in well with
the present analysis. They are, however, interesting in their own right, since they go counter
to previous results, which have shown an attenuation of the effects of aversive stimuli after
administration of alcohol (cf. Miller and
Barry, 1960).

REFERENCES
Azrin, N. H. Some effects of two intermittent schedules of immediate and non-immediate punishment.
J. Psychol., 1956, 42, 3-21.
Azrin, N. H., Holz, W. C., and Hake, D. F. Fixed-ratio
punishment. J. exp. Anal. Behav., 1963, 6, 141-148.
Brady, J. V. A comparative approach to the evaluation
of drug effects upon affective behavior. Ann. N.Y.
Acad. Sci., 1956, 64, 632-643.
Conger, J. J. The effects of alcohol on conflict behavior in the albino rat. Quart. J. Stud. Alcohol, 1951,
12, 1-29.
Fowler, H. and Miller, N. E. Facilitation and inhibition of runway performance by hind- and forepaw
shock of various intensities. J. comp. physiol. Psychol., 1963, 56, 801-805.
Geller, I. and Seifter, J. The effects of mneprobamate,
barbiturates, d-amphetamine and promazine on experimentally induced conflict in the rat. Psychopharmacologia, 1960, 1, 482-492.
Hunt, H. F. and Brady, J. V. Some effects of punishment and intercurrent "anxiety" on a simple operant. J. comp. physiol. Psychol., 1955, 48, 305-310.
Masserman, J. H. and Yum, K. S. An analysis of the
influence of alcohol on experimental neurosis in
cats. Psychosom. Med., 1946, 8, 36-52.
Miller, N. E. and Barry, Herbert III. Motivational effects of drugs: methods which illustrate some general problems in psychopharmacology. Psychopharmacologia, 1960, 1, 169-199.

Received August 21, 1963